They hop. They skip. And they jump. With frequent changes in speed and direction, they evade predators. They live a solitary existence that typically lasts six years—not bad for a little desert rodent.
Meet the jerboa, a tiny kangaroo-like creature that looks like it stepped right out of a sci-fi film. That’s apt, given that researchers at the University of Michigan believe the jerboa’s randomness of movement can be applied to the design of bipedal robots.
Talia Moore, a research fellow in ecology and evolutionary biology at U-M studying the jerboa, was at first frustrated by the animal’s lack of predictable patterns. “From step to step, it was making strategic decisions that you wouldn’t normally see in most animals,” added Ram Vasudevan, an assistant professor in mechanical engineering. Moore decided that it made more sense to study the jerboa’s unpredictability instead.
Working together, Moore and Vasudevan turned to information theory—which was formulated by U-M alumnus Claude Shannon, author of the landmark paper A Mathematical Theory of Communication, that introduced the concept of entropy, or uncertainty. Their study produced new optimization tools and mathematical ideas, as well as a new quantitative model for measuring randomness.
The new metric gives researchers the ability to create new biomimetic robots that can exhibit “natural” unpredictability in the way they move.
“This is something that’s not really been done in robotics before,” said Moore. “Mostly, we design robots to have this steady-state locomotion, and make continuous steps in a straight line.”
Noting that bipedal robotic systems might one day be deployed in environments such as deserts or the landscape of Mars, Vasudevan says that the study points to the benefit of exploiting the agility that unpredictability affords.
“If we want to make more biomimetic robots and think more about how robots can interact with nature and the world around us,” said Moore, “we’re going to need to think about how to engineer unpredictability in motion.”
Read more, from the University of Michigan: From rodent to robot
